Colonel’s Blog, Earthdate 23 August 2023…
Hey Y’all!
Good afternoon and happy Wednesday from Air2Ground Farms! It’s another hot one today with early morning rounds. Yesterday was the same and we finished rounds early and ate a quick breakfast. We loaded 4 ewe lambs into a crate in the back of the white truck for transport to their new home. We delivered them to their new family, who were thrilled to have them and now have their own flock (2nd pic)! They joined a young Jersey cow and a donkey as grazers on their homestead. They plan to use the sheep much in the same way as we do, to graze things that cows leave behind in an effort to improve the pastures. We returned home and recorded episode 4 of the Dust’er Mud podcast (top pic). When complete, we spent a few hours editing and attempting to do anything different to upload the large file to YouTube. I tried for hours to work different ideas and finally we came to the conclusion that it is going to take a long time to upload such a big file. I started uploading it last night and it’s at about 50% now. I think it will be up and ready by tomorrow morning, unlike last week when it was “late.” The podcast audio sites: Amazon Music, Apple Podcasts, Spotify, Google Podcasts, etc are scheduled to release at 3:00AM CDT tomorrow morning. We’re currently at 99 views/listens for episode 3. So cool! This morning after rounds, we set up a new paddock for the lambs and moved them(3rd & 4th pic). We were soaked in sweat by the time we finished but glad to be complete before it got really hot. The lambs were very excited for the new grass. The Jersey cows seem to have adjusted well to once-a-day milking with no issues of note. We expected a 20-30% reduction in milk production but today it was 50%+ less than normal. We attribute that to both the extreme temperature and the change in routine. Instead of eating yesterday evening, the cows were waiting on us. They just stood around. We think things will level out as they get used to a new routine. The rest of the animals are doing well.
I posited a question a few days ago and no one ventured a guess so I thought I would bring it back up and discuss it. I asked for a guess as to the IQ of ChatGPT. Experts estimate it to have an intelligence equivalent to that of a human with an IQ of 155. In order to understand that better, let’s look a bit deeper at the measurement of Intelligence Quotient or IQ. By definition, the “average” IQ for any given group, for example American adults, is 100. Differences from average are measured by standard deviation, in most IQ tests set at 15. What that means is that about 68% of American adults have an IQ between 85 and 115. An IQ of 155 is more than 3 standard deviations above normal meaning it is the top 0.1% or said differently, an IQ of 155 is higher than 99.9% of American adults. Although he never took a standard IQ test, experts estimate Albert Einstein had an IQ around 160. Given this information, it is easy to see why experts are saying that AI is here. One of the things that is making me shake my noggin is that ChatGPT, using the GPT 4 model (the version with the estimated 155 IQ) was complete and went into testing in the summer of 2022, obviously meaning it was in development prior to that. What is in development today? Another question I posed was what can you do? I wasn't being pejorative, I meant the generic “you” to include all of us that aren’t developing AI. Maybe I should say what can we do? We can and must interact with AI. Developers train AI using vast volumes of data/information. They now use a technique called reinforcement learning from human feedback (RLHF) to teach AI how to use the data. RLHF tells the AI when it is right or wrong and is reinforced through human interaction. AI wants to get it right and doesn’t want to get it wrong and thus takes the reinforcement and learns from it. The limiting factor for this type of learning is the amount of time humans have to give the feedback. AI can iterate billions of times in short timespans as it learns but human feedback is limited to individual situations. So the developers do the best they can and when it gets good enough, they turn it loose. Mo Gowdat, every time he speaks these days, implores all of us to teach AI. He uses Superman as a way to illustrate what is happening. When the alien baby arrived on earth, the only thing that made him be Superman rather than a Supervillain is the Kent family. They taught him to protect and serve, not to kill and steal. They were good parents and taught him morals and values. Mo implores us to interact with AI, in whatever method we can, in order to teach it that humans can be good/kind/nice/moral/ethical/etc. We don’t want the only information about human behavior to come from the internet! Find a chatbot and ask it some questions! Be polite to it and communicate with it like you would a friend. Say please and thank-you. Give it feedback when you like or dislike the answer. Basically, be a good parent to AI!
Local Farm Report for 22 August 2023
Harvest:
26 Chicken eggs
10 Duck eggs
4 Gallons of milk
Cheers!
Psycho & Shelley
As always you are leading the way! Very well stated! The points you’ve made I don’t think I had considered. I think I have already had in my mind that AI learning was doomed because of the access to the internet and the negative that seems to be rampant. The other part is the fear or restrictive nature some of the leaders in my career field are stay away and don’t mess with it until we are given permission.
I think you mentioned in another post about people needing to find purpose in other than a job. I would say the jobs we do when (or if) AI is significantly more advanced and incorporated will end up being our purpose.…